Student Prompting, Translanguaging, and Epistemic Motivation in LLM-Assisted Translation within EMI Contexts*

Published at American Journal of Applied Linguistics (AJAL)

​

Dr. L. H. Vogel, Ph. D. & Instructor

Germany

​​​

Abstract

The rapid integration of large language models (LLMs) into educational contexts is transforming both translation practices and multilingual learning environments. While previous research has examined student–LLM interaction in translation tasks and the role of translanguaging in English-medium instruction (EMI), limited attention has been paid to how these domains intersect. This study investigates how student translators’ prompting behaviors in LLM-assisted translation relate to translanguaging practices and epistemic motivation within EMI contexts. Adopting a mixed-methods design, the study analyzes interaction logs, translation outputs, and reflective responses from 40 Chinese EFL graduate students who completed Classical Chinese-to-English translation tasks using LLM support. A total of 168 prompts and corresponding outputs were systematically coded to identify prompting strategies, interaction patterns, and translanguaging behaviors. Additionally, post-task questionnaires and reflective journals were used to examine learners’ epistemic motivation and experiential perceptions. The findings indicate that students predominantly adopt a “generate–then–refine” prompting pattern, characterized by iterative engagement with model outputs. However, prompts targeting deeper semantic analysis, cultural adaptation, and pre-translation planning remain underrepresented. Quantitative analysis reveals that prompt frequency does not significantly predict translation quality, whereas the presence of human post-editing is strongly associated with improved outcomes. Translanguaging emerges as a key mediating process, enabling learners to draw upon multiple linguistic and cultural resources to interpret source texts and evaluate LLM outputs. From a motivational perspective, the study identifies a dynamic form of transepistemic motivation, characterized by the interplay between outward exploration (engagement with LLMs and multilingual resources) and inward reflection (critical evaluation and meaning construction). This dual process aligns with the “window and mirror” framework, illustrating how LLM-assisted translation environments function as spaces for both knowledge acquisition and epistemic self-positioning. The study contributes to the growing body of research on AI-assisted learning by integrating prompt literacy, translanguaging theory, and epistemic motivation into a unified analytical framework. Pedagogically, it highlights the importance of training students in strategic prompting, encouraging translanguaging practices, and fostering reflective engagement to enhance translation quality and learning outcomes.

Beyond Traditional Software Testing: A Cognitive and Prompt-Driven Framework for Large Language Model-Assisted Quality Assurance**

Published at Journal of American Academic Research (JAAR)

​

M. Atwood, PhD & Research Fellow

United Kingdom

​​​

Abstract

The rapid evolution of software engineering practices, driven by the widespread adoption of Large Language Models (LLMs), is fundamentally reshaping traditional paradigms of software testing and quality assurance. Conventional software testing methodologies—rooted in deterministic scripts, rule-based validation, and human-engineered test case design—are increasingly challenged by the scale, complexity, and dynamic nature of modern software systems. In response to these limitations, this study proposes a novel Cognitive and Prompt-Driven Framework for LLM-Assisted Software Testing, which reconceptualizes testing as an adaptive, semi-autonomous, and cognitively augmented process rather than a purely procedural engineering task. This research investigates how LLMs, when integrated with structured prompt engineering strategies and cognitive interaction models, can enhance software testing performance across multiple dimensions, including test coverage, defect detection efficiency, adaptability to requirement changes, and reduction in human cognitive load. Unlike prior studies that treat LLMs primarily as automation tools for test case generation, this work positions them as cognitive co-agents within a distributed testing intelligence system, capable of participating in reasoning, hypothesis generation, and iterative refinement cycles alongside human testers. The proposed framework is grounded in three theoretical pillars: (1) cognitive systems theory, which conceptualizes testing as a distributed cognitive process shared between humans and machines; (2) prompt engineering theory, which defines structured linguistic inputs as programmable interfaces for controlling LLM behavior; and (3) software testing optimization theory, which evaluates testing efficiency through measurable performance indicators such as defect detection rate, execution time, and coverage density. By integrating these perspectives, the study introduces a unified model that bridges human cognitive processes and machine-generated reasoning. Methodologically, the study adopts a quantitative-dominant mixed-methods experimental design, where traditional automated testing systems serve as the baseline control condition, while LLM-assisted testing frameworks constitute the experimental condition. A total of simulated software testing environments are constructed, representing diverse application domains including web systems, API-driven architectures, and enterprise-level distributed applications. Within these environments, multiple prompt engineering strategies—including instructional prompts, contextual prompts, iterative refinement prompts, and hybrid cognitive prompts—are systematically evaluated. Quantitative performance metrics are analyzed using comparative statistical techniques, including mean difference analysis, variance testing (ANOVA), and correlation modeling between prompt complexity and testing outcomes. In addition, qualitative analysis of interaction logs provides insight into human–AI cognitive collaboration patterns, revealing how testers adjust strategies in response to model outputs and how LLMs influence decision-making trajectories during testing cycles. Preliminary findings suggest that LLM-assisted testing significantly improves test case diversity and defect discovery rates, particularly when guided by structured contextual and iterative prompts. However, performance gains are highly dependent on prompt design quality, indicating that prompt engineering is not merely a supplementary technique but a core determinant of system effectiveness. Furthermore, the study identifies a phenomenon of cognitive redistribution, where human testers shift from low-level execution tasks toward higher-level reasoning, validation, and strategic oversight. The results also highlight critical limitations, including model hallucination risks, inconsistency in output generation, and dependency on prompt formulation expertise. These challenges underscore the necessity of integrating formal verification mechanisms and human-in-the-loop validation strategies within LLM-based testing systems. Overall, this study contributes a comprehensive theoretical and empirical foundation for next-generation software testing paradigms. It extends existing literature by framing LLM-assisted testing not as an incremental automation improvement but as a structural transformation in software quality assurance ecosystems. The proposed cognitive and prompt-driven framework offers both a conceptual model and an actionable methodology for designing intelligent, adaptive, and scalable testing systems in modern software engineering environments.

AI-Mediated Language Practices in Higher Education: Critical Perspectives on Literacy, Agency, and Discourse Transformation*

Published at American Journal of Higher Education (AJHE)

​​

Dr. M. A. Ribeiro, Ph.D.

Universality Associate Professor, Brazil

​

Abstract

The rapid integration of artificial intelligence (AI), particularly large language models (LLMs), into higher education is fundamentally reshaping how language is produced, interpreted, and evaluated in academic contexts. This study investigates AI-mediated language practices in higher education, focusing on the intersections of digital literacy, learner agency, and discourse transformation within contemporary applied linguistics. Unlike earlier computational approaches to language learning that emphasized rule-based or corpus-driven systems, generative AI introduces a dynamic, dialogic, and co-constructive linguistic environment in which meaning is continuously negotiated between human users and algorithmic systems. Drawing on critical applied linguistics, sociocultural theory, and digital literacy frameworks, this study conceptualizes AI not merely as a technological tool but as an active semiotic participant in academic discourse production. In this sense, AI-mediated communication is understood as a hybridized form of literacy practice in which students, instructors, and generative systems collaboratively construct texts, interpret meanings, and negotiate epistemic authority. The study argues that this shift represents not only a technological transition but also a paradigmatic transformation in how agency is distributed across human and non-human actors in educational settings. Methodologically, this research adopts a mixed-methods design combining large-scale survey data (N=420 students across higher education institutions), discourse analysis of AI-assisted writing samples, and interaction log analysis from AI-supported academic writing tasks. Quantitative findings are analyzed using regression modeling, factor analysis, and comparative variance testing, while qualitative data are examined through critical discourse analysis (CDA) and interactional sociolinguistics frameworks. The integration of these methods enables a multidimensional understanding of how AI shapes literacy practices across cognitive, linguistic, and institutional dimensions. Findings indicate three major patterns. First, AI-mediated literacy significantly enhances linguistic fluency and textual coherence, particularly in second-language academic writing contexts; however, it simultaneously introduces new forms of epistemic dependency, where learners increasingly rely on algorithmic suggestions for argument construction and lexical choice. Second, learner agency is reconfigured rather than diminished, shifting from independent text production toward strategic orchestration of AI-generated outputs. This reflects a transition from “author-as-producer” to “author-as-editorial curator,” suggesting a redefinition of academic authorship in digitally mediated environments. Third, discourse transformation occurs at both micro and macro levels: micro-level linguistic features such as cohesion markers, syntactic complexity, and lexical sophistication show measurable improvement, while macro-level academic discourse structures increasingly reflect homogenization patterns associated with AI-generated stylistic conventions. The study further identifies a critical tension between enhancement and dependency. While AI tools expand access to advanced linguistic resources and support multilingual learners in academic participation, they also risk narrowing discursive diversity and reducing the authenticity of learner voice. This tension highlights the need for pedagogical frameworks that integrate AI literacy as a core component of higher education curricula, emphasizing critical engagement, ethical awareness, and reflective use of generative systems. Theoretically, this research contributes to applied linguistics by extending sociocultural and digital literacy theories into AI-mediated environments, proposing a model of “distributed linguistic agency” in which meaning-making is co-constructed across human cognition and machine-generated discourse. Practically, the findings provide implications for curriculum design, academic writing instruction, and assessment practices in AI-integrated educational systems. Overall, this study positions AI-mediated language practices as a transformative force in higher education, reshaping not only how texts are produced but also how knowledge, authorship, and linguistic authority are conceptualized in contemporary academic discourse ecosystems.

Modeling Pragmatic Meaning in Digital Interaction: Integrating User Behavior, Contextual Dynamics, and Applied Linguistic Frameworks*

Published at American Journal of Applied Linguistics (AJAL)

​​

Dr. E. Whitcombe, Ph.D.
Assistant Professor of Applied Linguistics
USA

​

Abstract

The rapid expansion of digital communication environments has fundamentally reshaped how pragmatic meaning is constructed, interpreted, and negotiated. Traditional models of pragmatics, largely grounded in face-to-face interaction, are increasingly insufficient for capturing the layered, multimodal, and algorithmically mediated nature of contemporary discourse. This study proposes an integrative framework for modeling pragmatic meaning in digital interaction by combining user behavioral metrics, contextual dynamics, and applied linguistic theory. Drawing on a mixed-methods research design, the study examines how users construct meaning across multiple digital platforms, focusing on the interaction between linguistic form, contextual variability, and observable engagement patterns. The research is based on a dataset of 12,500 digital interactions collected from social media platforms, online forums, and messaging environments. These interactions were coded using a multi-level pragmatic annotation scheme, incorporating speech act classification, implicature density, politeness strategies, and contextual markers such as temporal sequencing and platform affordances. Quantitative analysis was conducted using multivariate regression, cluster analysis, and structural equation modeling to identify relationships between pragmatic features and user engagement outcomes. Findings indicate that pragmatic meaning in digital environments is not solely a function of linguistic encoding but emerges from the interaction between user intent, platform-specific constraints, and audience interpretation. High-engagement interactions were characterized by increased pragmatic layering, including indirectness, multimodal signaling, and adaptive politeness strategies. Conversely, interactions with lower engagement tended to rely on explicit, low-context communication forms. The study also demonstrates that contextual dynamics—such as thread structure, audience size, and temporal responsiveness—play a critical role in shaping pragmatic interpretation. These factors mediate the relationship between linguistic form and communicative effect, suggesting that pragmatic meaning must be understood as a dynamic system rather than a static property of utterances. The proposed model contributes to applied linguistics by offering a scalable framework for analyzing digital discourse, bridging the gap between theoretical pragmatics and computational approaches. It also has implications for digital communication design, suggesting that effective interaction strategies must account for both linguistic and behavioral dimensions. Ultimately, this research advances the understanding of pragmatics in the digital age by reconceptualizing meaning as an emergent property of interactional systems. Future research directions include the integration of machine learning models for automated pragmatic analysis and the exploration of cross-cultural variation in digital communication practices.

​

​

NOVEL LUBRICATE, KILL & BLEED METHOD**

Published at American Journal of Applied Mathematics (AJAM)

Dieudonne NDONG OVONO, Ph.D
TotalEnergies OneTech, France

​

Abstract

This paper investigates the application and limitations of the so-called “Lubricate & Bleed” (L&B) method — a volumetric technique widely used to address oil well integrity challenges. In practice, significant discrepancies were observed between the predicted and actual fluid pumping and purging schedules when applying this method in the field. This study examines the scientific principles underpinning the L&B method and proposes a refined predictive framework to improve its accuracy and reliability.

The L&B process involves a sequence of fluid injection (lubrication) and controlled venting (bleeding), with operational steps typically determined using standard procedures found in professional well control manuals. However, field experience has shown that the actual execution of these steps often deviates markedly from the initial plan. In several operations, it was necessary to manually adjust the fluid volumes of specific steps to achieve a successful outcome. These discrepancies between predicted and actual performance motivated a deeper investigation into the method’s theoretical foundations, with the aim of enhancing future operational practices.

To illustrate the core challenge, consider the analogy of a closed bottle containing gas that must be entirely replaced with water — without breaking the bottle or releasing the gas into the atmosphere. Even with a valve allowing connection to external flow control devices, completely displacing the gas is non-trivial due to gas expansion and distribution effects. When this analogy is extended to a wellbore — a vertical closed space extending hundreds or thousands of meters — the challenge becomes even more pronounced.

Nevertheless, displacing the gas with a safe liquid such as water remains achievable through the lubricate and bleed technique. This paper first reviews the conventional L&B methodology described in well control manuals. However, during field applications, the method proved less effective than anticipated due to fundamental limitations in the underlying assumptions and calculations.

Moreover, additional challenges arose when the well exhibited sustained casing pressure (SCP), revealing critical limitations of the standard L&B approach. This prompted the development of a modified strategy — the “Lubricate, Kill & Bleed” (LK&B) method — which not only displaces the gas but also allows for effective well killing operations.

Building on the thermodynamic behavior of real gases, equations of state, echometry, laboratory analysis of gas properties, calculus of variations, partial differential equations, energy conservation laws, heat transfer models, and product series expansions, this study derives an improved predictive model for volumetric well control operations. The proposed framework offers enhanced accuracy in forecasting fluid schedules and provides operational guidance for complex well integrity scenarios involving trapped gas under pressure.

Positive and Negative Benefits of Flaxseed as an Organic Food Source During the Period of Sustainable and Strategic Development **

Published at American Academic Journal of Agriculture (AAJA)

​

Süleyman Özberk, Ph. D. and Lecturer
Cukurova University, Turkey

​

Abstract

Flaxseed was cultivated in Babylon in around many centuries ago and many people consider it as one of the most powerful plant foods on the planet (Magee, 2020). When Traced back to history, we can also realize that flaxseed can be used to treat diseases such as heart disease, diabetes, stroke and all kinds of cancer. The application of flaxseed was highlighted by scientists and medical workers for many centuries. However, the valuable effects of flaxseed was always underestimated to a large degree, and even people who are working in the specific field can not cognitively associate the benefits of flaxseed with the treatment of certain disease. Thus, the waste of flaxseed was unnoticed when it comes over-consumption and non-effective consumption. This article will enumerate both the positive side and the negative side of flaxseed which will benefits sustainable development as an organic food source. After a systematic comparison and analysis, the conclusion was drawn naturally with statistic analysis and logic induction as scientific collaboration. *

​

​

From Western Culture to Eastern Culture: A Dialogic Perspective on the Holistic Overview of Chinese Culture and American Culture **

​

Published at Journal of American Academic Research (JAAR)

​

Wenbin Xue, Ph. D. & Lecturer
China University of Mining & Technology, China

​

Abstract

Due to globalization, the clear-cut contrast between Chinese culture and American culture was blurred through constant cultural dialogue and interaction. Culture is complex and multidimensional. It is in fact too complex to define a certain culture in simple terms and rough definition. Kroeber and Kluckhohn in 1952 identified over 160 different definitions of culture. One of the earliest widely cited definitions by Tylor in 1887 defines culture as “that complex whole which includes knowledge, belief, art, morals, law, custom, and any other capabilities and habits acquired by human as a member of society.” With a dialogic perspective, this article investigated the differences and common features between western culture and eastern culture by observing the holistic overview of specific culture in China and in America. The research result confirms our hypothesis. The research also suggested that people in modern era should view distinct culture with more tolerance, appreciation and respect. New culture should be redefined with respect to historical facts, linguistic uniqueness and differences in customs.  *

​